A novel fault diagnosis method for satellite systems based on multivalue logic inference with semi-tensor product of matrices

Abstract

Due to complexities in satellite systems, the relationship between faults and symptoms become more complex, usually each fault may be related to several fault symptoms and some different faults can result in the same symptoms, so as to make the fault diagnosis very hard with direct symptom analysis. Some suspected faults could be found through symptom analysis combined with a priori knowledge, however it is very important to achieve effective diagnosis with a series of uncertain and unreliable factors. This paper presents a novel fault diagnosis method for satellite systems based on multivalue logic inference with semi-tensor product of matrices. Firstly, the multivalue logic methodology is employed to define fault symptoms and conditions as propositions thus the logical values of various propositions can be expressed as vectors. In this way, the multivalue logic inference is implemented by some fast operations with the semi-tensor product of matrices which extends matrix product operations to any case of incompatible dimension matrices. Then the probabilities of suspected faults can be found through mathematical calculation of multivalue logic, according to which the diagnosis decision is made for various faults and the corresponding spare switch is carried out to guarantee the online safety of satellite systems. Finally the feasibility and effectiveness of the proposed method are verified and validated by some simulation experiment results and its advantages are also remarked briefly.